HY-1C卫星COCTS近海水体遥感反射率产品真实性检验

Validations of the HY-1C COCTS remote sensing reflectance products in coastal waters

中国HY-1C卫星水色水温扫描仪(COCTS)从2018年9月发射至今已积累了大量的全球观测数据,本文利用AERONET-OC实测数据集对COCTS近海水体遥感反射率(Rrs)产品进行了真实性检验。本文对实测光谱数据集进行归一化处理,按光谱特征分为A (清洁)、B (比较清洁)、C (轻微浑浊)、D (比较浑浊)这4种光学水体类型,对COCTS Rrs产品在这4类水体的精度进行评价。结果表明:COCTS反演Rrs在A、B两种水体类型中存在轻微高估(平均绝对百分比误差APD (Average absolute Percent Difference)分别为38.79%、44.44%),在C类水体轻微低估(APD=40.85%),而在D类水体显著低估(APD=47.14%)。COCTS不同波段Rrs产品在4类水体的精度亦存在差异,其中412 nm、443 nm波段在A类水体中和实测数据一致性相对较好(APD均小于30%);520 nm、565 nm波段在C、D两类水体中和实测数据一致性相对较好(APD均小于30%);670 nm波段在4种水体均呈现显著低估。由于AERONET-OC波段和COCTS存在差异,可能会导致评估误差存在高估,未来需要使用连续光谱的实测数据进一步评价COCTS产品精度。

A large number of global observation data was obtained by HY-1C COCTS since its launch in September 2018. The comprehensive evaluation of the HY-1C/COCTS products is important for further applications. In this study, we used the global in-situ data from AERONET-OC to evaluate the performance of the remote sensing reflectance(Rrs) products of the HY-1C COCTS. Firstly, the AERONET-OC dataset were divided into four optical water types(A, clean water;B, relatively clean water;C, slightly turbid water;D,turbid water) based on a spectral normalization method. Secondly, the AERONET-OC Rrs data and HY-1C/COCTS retrieved Rrs data were matched according to the defined spatial-temporal windows(5×5 box and 1 hour). Finally, the performances of the HY-1C/COCTS Rrs products were quantitatively evaluated in the four optical water types. As a result, good correlation between satellite and in-situ Rrs data was indicated as R values among four types water ranged from 0.680 to 0.879. In type A water, the relatively good consistency between satellite and in-situ Rrs data was observed as average percent difference(PD) at 6.79%, and the average absolute percent difference(APD) at 38.79%. In type B water, slight overestimation of satellite data occurred with PD at 18.73% and APD less than 45%. Underestimation of satellite data was reported in type C water, as negative remote sensing Rrs data in 412 nm and 443 nm bands were with PD at-14.38% and APD at 47.14%. Similarly, in type D water, negative Rrs in 412 nm and 443 nm bands were with PD at-32.35%, and APD at 47.14%,indicating significant underestimation from the satellite data. In addition, difference of accuracy performance of Rrs products in different bands of COCTS for four water types was also observed. Rrs presented good consistency between in situ and COCTS data in band 412 nm and 443 nm for type A water. Better consistency in band 520 nm and 565 nm was observed for type C, D water than type A water while significant underestimation of COCTS Rrs were reported in all four types of water compared to in situ data. Overall, COCTS and in situ Rrs data showed good consistency in clean water, but remained relatively inconsistent in turbid water. Our results also reported that the COCTS inversed Rrs products are slightly overestimated compared with the AERONET-OC in situ data in type A and B water. In contrast, COCTS products slightly underestimated Rrs for type C water but significantly underestimated for type D water. Attention should also be paid to enlarge the evaluated errors, as the AERONET-OC in-situ spectral data was linearly interpolated to match COCTS band. In the future, the hyperspectral in-situ Rrs data should be used to further evaluate the performance of COCTS.